About This Software

TURN 2D IMAGES INTO 3D BASE MESHES IN MINUTES. Jupetar is a powerful, locally-hosted generative 3D pipeline. Built on a heavily customized, cutting-edge multi-model architecture, it transforms a single 2D image—a concept art sketch, a photo, or a texture—into a fully textured, 4K-upscaled 3D model to kickstart your game development workflow.

Forget monthly subscriptions, cloud queues, and data privacy risks. Developed by Odyssey Game Studios, Jupetar runs 100% locally on your hardware. You own the software, you own the models, and your unreleased assets never leave your machine.

EXPECTATION MANAGEMENT: WHAT JUPETAR IS (AND ISN'T) Jupetar is designed to be a massive workflow accelerator, not a magic bullet. The models generated are base meshes. While V2.0 introduces native 4K upscaling and physically-based material injection, saving you hours of manual block-out time, the raw exports are not strictly "game-ready" out of the box. You will encounter minor artificing or topology that requires a quick cleanup pass in Blender before dropping it into your game engine.

V2.0 CORE FEATURES

• ⚡ The NTC C++ Bridge & Dynamic Hoister: Jupetar V2.0 completely rewrites how Python interacts with your GPU. By dynamically hoisting massive neural networks in and out of your VRAM and bypassing PyTorch bottlenecks, generation times have been slashed to under 10 minutes on a standard 10GB RTX 3080.

• 🧠 Deep VLM Anchoring: The engine now features a local Vision-Language Model (VLM) that analyzes your 2D image before generation begins. It automatically writes hyper-detailed prompt engineering directly into the core diffusion layers to ensure your models are generated with physically accurate materials (e.g., sharp metal instead of "plush plastic").

• 🎨 True SyncMVD Texturing: The texture engine doesn't just project the image; it paints your model from 6 virtual cameras simultaneously. Seams are perfectly stitched, and blind spots (like the bottoms of feet) are automatically hallucinated and filled.

• ✨ Zero-VRAM 4K Refinement: After generation, Jupetar utilizes an integrated Real-ESRGAN neural upscaler paired with CPU-bound OpenCV mathematics (Laplacian sharpening and micro-noise injection) to transform 512px textures into crisp, gritty, 4K PBR Albedo and Roughness maps—costing your GPU 0 extra VRAM.

• 🛠️ Workflow-Ready Output: Exports directly to .GLB, .OBJ, and .PLY formats. Ready to be imported straight into Blender for final topology cleanup before heading to Unreal Engine 5, Unity, or Godot.

DESIGNED FOR:

• Indie Developers: Populate entire levels with background props and clutter in a single afternoon.

• Modders: Turn 2D sprites from classic games into 3D objects for modern remakes.

• Concept Artists: Rapidly prototype 3D block-outs from your 2D sketches to verify forms and lighting.

ROADMAP: FUTURE UPDATES Jupetar is actively evolving. Post-launch implementations currently in development:

• AMD GPU Compatibility: Active development is currently underway to expand our hardware support and bring the full Jupetar generation pipeline to AMD Radeon graphics cards.

• Art Style Controls: UI toggles to easily switch the engine's internal mathematical refinement between "Realistic / Hard-Surface" and "Stylized / Low-Poly."

• Quality of Life (QOL) Updates: Expanded export formats, custom lighting controls, and further pipeline optimizations.

⚠️ STRICT SYSTEM REQUIREMENTS WARNING ⚠️ Jupetar is a highly demanding local AI application. Because it loads massive neural network weights directly into your graphics card, VRAM is an absolute requirement. Please review carefully before purchasing to ensure your system can handle the pipeline.

• Minimum (Slower Generation): NVIDIA GPU with exactly 10GB of VRAM (e.g., RTX 3080 10GB, RTX 3060 12GB, RTX 2080 Ti 11GB).

• Recommended: NVIDIA GPU with 12GB to 16GB+ VRAM (e.g., RTX 4070, RTX 4080, RTX 3090, RTX 4090).

• Not Supported: AMD Radeon (currently in development), Intel ARC, Integrated Graphics, or any NVIDIA GPU with 8GB of VRAM or less.

AI Generated Content Disclosure

The developers describe how their game uses AI Generated Content like this:

Jupetar is a dedicated 2D-to-3D asset generation engine powered entirely by generative artificial intelligence. The software utilizes localized, open-weights AI networks (including diffusion and 3D generation models) running directly on the user's hardware to convert user-provided 2D concept art and images into 3D meshes and textures. All AI processing is performed offline on your local GPU. Jupetar does not connect to external cloud generation services, and your personal assets, images, and prompts are never uploaded, collected, or used to train external AI models.